diff --git a/src/utils/process_data.py b/src/utils/process_data.py
index dff56b5..555cf0b 100644
--- a/src/utils/process_data.py
+++ b/src/utils/process_data.py
@@ -1,38 +1,39 @@
-# utils/process_data.py
-import json
 import sys
+
+from utils.handlers import get_build_handler
 from .paths import get_project_path
 from sacrebleu import sentence_bleu as bleu
+from utils.dataset import ArchiveState, Dataset
 
+REFERENCE_MAP = Dataset.from_json(
+    str(get_project_path('../data/dataset.json'))
+).build_reference_map()
 
-def build_reference_map(dataset_path: str) -> dict[str, list[str]]:
-    ref_map = {}
-    data = json.loads(open(dataset_path).read())
-    for entry in data['entries']:
-        id_ = entry['metadata']['id']
-        comments = entry['comments']
-        ref_map[id_] = [c['body'] for c in comments]
-    return ref_map
-
-
-REFERENCE_MAP = build_reference_map(str(get_project_path('../data/dataset.json')))
+ARCHIVES_ROOT = str(get_project_path('../data/archives'))
 
 
 def evaluate_comments(answers: dict[str, str], percent_cb):
     total = len(answers)
     results = {}
+    print(REFERENCE_MAP.keys())
     for i, (id_, gen) in enumerate(answers.items(), 1):
         if id_ not in REFERENCE_MAP:
             print(f'id: "{id_}" is not present in the dataset', file=sys.stderr)
             continue
-        paraphrases = REFERENCE_MAP[id_]
-        max_score = 0.0
+        entry = REFERENCE_MAP[id_]
+        max_score = 0
         scores = []
-        for p in paraphrases:
-            score = bleu(gen, [p]).score
+        for p in [entry.comments[0].body] + entry.comments[0].paraphrases:
+            score = round(bleu(gen, [p]).score, 2)
             scores.append(score)
             max_score = max(max_score, score)
-        results[id_] = {'max_bleu_score': max_score, 'bleu_scores': scores, 'proposed_comment': gen}
+
+        print(scores)
+        results[id_] = {
+            'max_bleu_score': max_score,
+            'bleu_scores': scores,
+            'proposed_comment': gen,
+        }
         percent_cb(int(i / total * 100))
     return results